The FeatureReader and FeatureWriter are ideal candidates to allow concurrent data reading and writing, something that you can't do with multiple readers and writers right now. Adding these parameters is currently possible if you wrap them up in custom transformers like I explain in this webinar, but ideally you shouldn't have to go to this effort.
https://www.youtube.com/embed/YnCot12FA9ULooking forward to the performance gains this change would bring. In tests my timings suggest in some examples it reduces the readwrite times to a third or even less.
Cheers, Dave
Anything that can speed up reading and writing will get my vote!
Not sure though how this can apply to file based datasets like DWG, DGN and GML.
I know FME 2019 has pushed parallel processing down into Custom Transformers, but when I'm loading-up two 50K+ datasets to compare, I'd much like to load them in parallel, rather than sequentially.
I guess concurrent reading of separate files would be an advance. Ideally each stream in the workspace should run concurrently, up to some process maximum.
I had to do a spatial filter on a very large data set and was a bit surprised that the feature reader doesn't do it asynchronously. Gave up after a while and had to solve it in Python. 😞